retro-dv-home-video-by-2zdsi5
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill relies on ingesting content from a local file that contains external prompt excerpts and metadata, which the agent uses to generate output.
- Ingestion points: The file
references/cases.mdis read during every invocation of the skill to select an anchor case. - Boundary markers: The instructions lack explicit delimiters or 'ignore embedded instructions' warnings for the data read from the reference file.
- Capability inventory: The skill instructs the agent to use available image, video, or browser tools to inspect media at external URLs.
- Sanitization: There is no evidence of sanitization or filtering applied to the external prompt excerpts or the content of the linked media pages.
- [EXTERNAL_DOWNLOADS]: The skill directs the agent to interact with external domains to verify media quality.
- Evidence:
references/cases.mdcontains multiple links togoodcase.aiandmedia.goodcase.aifor video and image inspection. While these are contextually relevant to the skill's purpose, they are non-standard domains. - Evidence: The skill also references
x.com, which is recognized as a well-known service.
Audit Metadata